[Standards] [Fwd: Re: Namespace well-formedness and RFC3920bis backwards compatibility]
dave at cridland.net
Tue Sep 23 18:46:56 UTC 2008
On Tue Sep 23 18:23:45 2008, Jonathan Schleifer wrote:
> Sorry, I didn't read the whole thread as I'm on vacation, but it
> seems to me you are talking about the routing of unbound
> namespaces - a problem we have with Gajim at the moment. When we
> receive an unbound namespace, our XML parser will just fail and
> there's not much we can do about that. It's nearly impossible to
> fix for us, it would require changes to Python and pyxmmp, if I
> looked correctly.
No, that's not the case, I stuck a patch on #3083 that seems to
almost work. It's working enough for me to be running it anyway.
The trick in this instance was to tell the XML parser not to bother
handling xml namespaces, and do it ourselves, giving considerably
more flexibility. I'd be willing to bet that in most cases, it'd be
easier - Gajim's a bad case because xmpppy itself has rather bizarre
handling of namespaces to begin with.
> OTOH, it's easy to not route unbound namespaces.
Not true. It's a substantial performance impact for us, and
worthwhile avoiding in many instances.
> There is already a patch for ejabberd that does this. The
> performance loss introduced by it is so minimal that you can
> ignore it.
That's largely because ejabberd already dopes much deeper parsing
than for example we do. And in any case, this does not mean that
Gajim doesn't have to bother solving its own issues.
> There is a ticket for that bug, which I unfortunately don't have
> handy atm. And I can't look it up as I'm writing the message
Dave Cridland - mailto:dave at cridland.net - xmpp:dwd at dave.cridland.net
Infotrope Polymer - ACAP, IMAP, ESMTP, and Lemonade
More information about the Standards